Salinity & Temperature

Data assimilation

N. HOAREAU (ICM, SMOS-BEC)

M. UMBERT (ICM)

S. KALARONI (ICM, SMOS-BEC)

J. BALLABRERA (UTM, SMOS-BEC)

2010/12/15

SUMMARY

• 1. Region of interest

• 2. NEMO-OPA

• 3. ARGO data

• 4. NUDGING, results …

• 5. Perspectives…

1. Region of interest

A region of validation for SMOS data …

45W 2ºW

44ºN

15ºN

• PE Model: NEMO-OPA

• 1/3º horizontal resolution, 31 vertical levels.

- Partial steps (better topography resolution)

- Zero Eddy Induced Velocity (development of turbulence)

• Simulation period: 2000, Jan 1st – 2009, Dec 31.

• Spin up: 15 years simulation from Levitus, at rest, and climatological forcing (Dr. Baptiste Mourre, MIDAS-4 and 5).

• Open boundary conditions, seasonal data (MERCATOR).

• Atmospheric forcing (NCEP-NCAR):

- DAILY: Wind stress, 10m Wind speed, 2m Air temperature

- MONTHLY: Precipitation rate, Cloud cover and Humidity

2. NEMO-OPA : model

Regular Z coordinate Partial-step Z coordinate

2. NEMO-OPA : Mean value (SST, SSS)

Temperature gradient SW-NE

Max=26ºC, Min=16ºC

Upwelling off the coast

Tongue of salty water in SW

Max=37.5 ; Min=35

Strong meridional gradient

2. NEMO-OPA : Variability (variance) (1/2)

Region of low variability in the South

Region of higher variability in the North (seasonal cycle)

SSS variance < 0.01

2. NEMO-OPA : Variability (SST,SSS EOFs) (2/2)

Principal component (mode1)

3. ARGO dataCumulated ARGO floats2002/01/01 - 2009/12/31

Coriolis database24 047 ARGO profiles

14 438 ARGO profiles

Pre-process and selection

1- Not in the “Grey list” (a list of known wrong ARGO)

2- Good QC for T, S and P

3- Departure from Levitus < 5ºC for T ; < 2psu for S

3. ARGO data: EOF fitting

3D-Interpolation onto

NEMO grid by multivariate

EOF fitting:

- 10 EOFs, which represent

61% of the total variability.

- Monthly field

CROSS-VALIDATION: RMS (Argo original – Argo interpolated), as a function of depth and number of EOFs.

Temperature Salinity

4.Nudging Method

•Relaxation term is added into the equation of evolution of a prognostic variable (in our case, T and S).

•The nudging term tends to reduce exponentially the distance of the model towards the observations.

0dxPhysics x x

dtRelaxation coefficient (s-1)Prognostic variable T, S, U,

V.

Observations

• Reynolds SST data

Adjusted manually

• 3D fields of T and S (ARGO data)

we use the relationship relating the nudging coefficient with the expected error of the observation:

1-

22

2

s *1

)(

)(rmsC

C

ztz

)days 10( 10 when 1042478.6 -11-642

C s

)1day( s 10 -1-15

4.Nudging: coefficient

Ref: J. BALLABRERA (PhD thesis, 1998)

Temperature

28 of January 2002

4. Nudging SST(x,y) + T(x,y,z) + S(x,y,z)

Salinity profile

Salinity profile

Salinity profile

28 of January 2002

• Reduction of fresh water extension at 36ºN.

• Weaker salty tongue.

• Inside the domain, variability comes from observations.

• At boundaries, variability comes from model (open boundary data).

Nudging validated

5. Perspectives …

• Investigate added value of assimilation of SSS from:

- ARGO data (new product from OA)

- SMOS data

• Use other methods of data assimilation,

as the Ensemble Kalman Filter (EnKF)

Thank you !!!